Mobile IPs Are a Different Product: A Signal Report on 4G/5G vs. Residential vs. Datacenter

Proxy buyers treat “residential” and “mobile” as price tiers of the same thing. They are different products: to anti-bot systems, a mobile IP announces “a phone on a carrier network,” and that signal changes what you’re served, what blocks you, and what your data is worth. This report maps the three IP classes to the workloads where each is the correct answer — with the published prices attached.

Think of it as a signal report rather than a buyer’s guide: the question isn’t “which IP is best” but “what does each IP class tell the target site about who is asking.”

The Three Signals

Datacenter: “I am a server.” Fast, cheap, honest. Perfect for targets that don’t discriminate (public APIs, sites without bot defenses, your own uptime checks) and useless against anything that scores IP reputation, because datacenter ranges are the first thing every blocklist contains.

Residential: “I am a home connection.” The workhorse signal — 100M+ IPs across 190+ countries on Thordata’s published pool, with country/state/city/ASN targeting and sticky sessions up to 90 minutes, priced on a volume slider ($2.00/GB at 1 GB down to $0.65/GB at multi-TB scale). Most collection workloads should default here.

Mobile: “I am a phone on a carrier network.” The rarest and most interesting signal. Carrier-grade NAT means thousands of real users share each mobile IP range, so blocking a mobile IP means blocking real customers — targets are structurally more reluctant to block them. Thordata’s mobile pool runs roughly 600,000 4G/5G IPs, priced from $5.00/GB at low volume down to $2.20/GB at scale.

The Signal-to-Workload Matrix

The matrix’s honest reading: mobile is a specialist tool. Paying $2.20–5.00/GB for traffic that would succeed on a $0.73/GB residential session is waste; paying residential prices for a mobile-gated endpoint that fails every time is worse waste. The failure mode teams hit most is not knowing which column they’re in.

Testing Which Column You’re In

Before committing volume, run the discriminator test — same request, three IP classes:

import requests

ENDPOINTS = {
    "datacenter": "http://USER:PASS@dc.thordata.com:18000",
    "residential": "http://USER-geo-us:PASS@res.thordata.com:18000",
    "mobile": "http://USER-geo-us:PASS@m.thordata.com:18000",
}

def signal_test(url):
    results = {}
    for name, ep in ENDPOINTS.items():
        try:
            r = requests.get(url, proxies={"http": ep}, timeout=30)
            results[name] = {"status": r.status_code,
                             "serves_full_content": looks_complete(r.text)}
        except Exception as e:
            results[name] = {"status": "error", "error": str(e)}
    return results

# Run 20x per target per day for a week; the pattern is the answer

Three possible readings: all three succeed (use datacenter, save the budget), only residential and mobile succeed (default residential), or only mobile succeeds (you’ve found a mobile-gated target — mobile is now the product, not an upgrade). The week of testing costs a few dollars and prevents a quarter of mispriced traffic.

The Economics, Straight

At published rates, the decision boundary is clean. Residential at $0.65–2.00/GB handles general collection. Mobile at $2.20–5.00/GB is justified only when (a) the discriminator test shows the target is mobile-gated, or (b) the workload is inherently mobile — app-store collection, mobile ad verification, mobile-authentic SERP collection. Datacenter at $0.75 per IP (Thordata’s published static pricing, with ISP at the same per-IP rate) covers the unguarded tier.

One crossover worth knowing: for mobile search results specifically, per-response SERP pricing can beat both proxy classes. A structured mobile SERP response from Thordata’s SERP API runs about $0.70–$1.20 per 1,000 responses — you’re paying for the structured answer, not the transport, and the SERP monitoring solution packages that into scheduled mobile rank tracking. If the workload is “know the mobile SERP,” that’s the product; if it’s “emulate a mobile user through a flow,” mobile IPs are.

FAQ

Why would a site serve different content to mobile IPs?
For the same reasons it serves different content to mobile devices: app-download prompts, mobile-only layouts, carrier-specific localization, and fraud rules that trust carrier traffic more. If your data model assumes one web per URL, mobile-gated surfaces will quietly disagree.

Is mobile “safer” for ToS-sensitive work?
No IP class makes a use case appropriate — policy comes from the target’s terms and your compliance review, not the transport. What mobile changes is structural blocking economics: shared carrier NAT means blocking mobile IPs costs targets real users. That’s an observation about reliability, not permission.

Can I mix mobile and residential in one session pool?
Yes, and the sticky-session design is the same (up to 90 minutes on Thordata), but don’t mix within a session — a session whose IP jumps from residential to mid-flow looks like exactly what it is. Split pools by workload, not by request.

How do mobile IPs relate to mobile SERP data?
They’re the emulation path; structured SERP is the data path. Emulation matters when you’re testing flows; when the deliverable is rank or content data, continuous SERP data crawling with device-level results is cheaper and cleaner than scraping through proxies. Most teams use both: proxies for the flows, structured SERP monitoring for the data.

The report’s closing recommendation is the same as its opening claim: stop treating IP classes as price tiers and start treating them as signals. Run the discriminator test on your top ten targets, price each workload in its correct column, and let the matrix — not the marketing — decide where your budget goes. Free trial credits cover the test week on every class.